Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Contents lists available at ScienceDirect

Spectrochimica Acta Part A: Molecular and

Biomolecular Spectroscopy
journal homepage: www.elsevier.com/locate/saa

Multi-species hydrocarbon measurement using TDLAS with a wide

scanning range DFG laser
Qiming Wang a, Zhenzhen Wang a,b, Takahiro Kamimoto a, Yoshihiro Deguchi a,b,⇑, Shengli Cao a,c,
Du Wen a,b, Daichi Takahara a
a
Graduate School of Advanced Technology and Science, Tokushima University, Tokushima 770-8501, Japan
b
State Key Laboratory of Multiphase Flow in Power Engineering, Xi’an Jiaotong University, Xi’an 710049, China
c
School of Power and Energy Engineering, Xi’an Jiaotong University, Xi’an 710049, China

h i g h l i g h t s g r a p h i c a l a b s t r a c t

A DFG laser system with a wide

spectral range was proposed on
hydrocarbon measurement.

Fifteen hydrocarbon spectra from C1
to C8 were measured.

Concentration dependence of
absorbance for each hydrocarbon was
evaluated.

Binary to trinary mixtures were
analyzed using the ICA and the CLS
method.

a r t i c l e i n f o a b s t r a c t

Article history: Tunable diode laser absorption spectroscopy (TDLAS) is a widely used diagnostic technique due to its
Received 12 April 2021 high sensitivity, fast response, low cost, and other merits. Hydrocarbon detection is a field of great inter-
Received in revised form 10 August 2021 est in the application of tunable diode lasers as hydrocarbons are fundamental molecules in many indus-
Accepted 24 August 2021
trial processes. Many tunable diode lasers are only suitable for single species detection due to the short
Available online 28 August 2021
scanning range and in real situations. However, different hydrocarbon species tend to exist simultane-
ously. Here we present a laser system based on the difference-frequency generation (DFG) method for
Keywords:
simultaneous hydrocarbon mixtures detection. The direct absorption spectra of different hydrocarbons
Hydrocarbon measurement
Tunable diode laser absorption
covering various groups (e.g., alkane, olefin, and aromatic) were measured. The measurements of the con-
spectroscopy (TDLAS) centration dependence of absorbance for each molecule were carried out. The R2 values were larger than
difference frequency generation (DFG) 0.997, which demonstrated the system can measure hydrocarbons covering different molecular classes
accurately. The mixture components were identified using the independent component analysis and
quantitative analysis was performed using the classical least-squares method. Future studies will focus
on the validation of the system in actual processes.
Ó 2021 Elsevier B.V. All rights reserved.

1. Introduction

Tunable diode laser absorption spectroscopy has been used for

gas sensing for a long time [1]. With the development of laser
sources and electro-optical components, TDLAS has turned into a
⇑ Corresponding author at: Graduate School of Advanced Technology and Science, mature diagnostic method and is applied to various fields [2–4].
Tokushima University. 2-1, Minamijyosanjima, Tokushima 770-8506, Japan It has the merits of high sensitivity, fast response, reasonable costs
E-mail address: ydeguchi@tokushima-u.ac.jp (Y. Deguchi).

https://doi.org/10.1016/j.saa.2021.120333
1386-1425/Ó 2021 Elsevier B.V. All rights reserved.
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

compare to other traditional diagnostic techniques like gas chro- is path length, Si;j is absorption line strength of absorption line j and
matography [5] and Fourier transform infrared (FTIR) spectrome- Gv i;j is line broadening function.
ters [6]. Hydrocarbon detection has become an important field The independent component analysis (ICA) is a signal process-
for the application of laser absorption spectroscopy [7–9]. As the ing method for signal separation and has been applied for chemo-
fundamental ingredients (either intermediate or final) or the prod- metrics usage (mixture decomposition) by many researchers [33–
ucts of chemical processes like coal pyrolysis and combustion, mix- 35]. The linear ICA problem can be expressed as:
tures of different hydrocarbons will appear simultaneously. The
X ¼ AS ð2Þ
detection of various hydrocarbon gases is necessary for the clarifi-
cation of chemical reactions as well as the control of industrial where X is an M  N matrix of the measured mixture spectra, S is a
processes. K  N matrix of K unknown spectra of pure components, and A is an
Hydrocarbon species contain a C-H bond that has a fundamen- M  K mixture matrix (unknown concentration). The ICA approach
tal vibrational mode. The C-H stretch results in many absorption tries to find the optimal separation matrix W to make the compo-
bands in 3 lm and 10 lm, making these regions easy for hydrocar- nents of Y as independent as possible.
bon measurement. Many previous works that measure hydrocar-
bons fall into the near-infrared to mid-infrared region (2.5– Y ¼ MX ¼ W T X ð3Þ
12 lm). Detection of methane [10], ethane [11], and propane In this work, the FastICA [36] algorithm was applied for compo-
[12] were performed using various DFB lasers. Quantum cascade nents identification due to its fast speed.
lasers (QCL) and interband cascade lasers (ICL) provided new To determine the concentrations for each component in a mix-
opportunities for hydrocarbon detection [13–17]. Laser sources ture, the classical least-squares (CLS) method was applied. It is a
based on difference-frequency generation (DFG) were also widely well-known regression method and can be applied directly to
discussed and applied to hydrocarbon detection with the advances quantitative spectral analysis.
in optical materials, lasers, and photonics technology [18–20]. The Another way of representing the Beer-Lambert law for multiple
DFG lasers can achieve wide and rapid tunability, and a lot of components at a single wavelength is given by the following
research based on DFG lasers have been reported on various hydro- equation:
carbon detection like methane [21–24], ethene [25], ethane
X
n
[26,27], propylene [28], benzene [29,30], and some other hydrocar- A¼ ki c i ð4Þ
bon fuel components [31,32]. i
Despite the benefits and wide application, most TDLAS tech-
where A is the absorbance of the mixture at a given wavelength, ki
niques are designed to measure specific target gas due to their nar-
is the absorbance coefficient and ci is the concentration of the pure
row spectral range limitation. Also, spectrally isolated features
substance. By expanding this formula to multiple wavelengths, the
were required for quantitative measurements thus need fine-
following matrix equation can be obtained:
tuning for the desired wavelength. Previous reports on hydrocar-
bon measurement using DFG lasers share the same limitation with A ¼ KC ð5Þ
other tunable diode lasers and mainly focus on the measurement
where A is the absorbance matrix of the mixture, K is a matrix of
of single species. The limitation on the spectral range and simulta-
the pure component spectra making up the mixture, and C is a sin-
neous measurable hydrocarbon species poses obstacles to the fur-
gle column matrix of concentration. The idea of this method is to
ther application of TDLAS.
find the concentration for each component that minimizes the
Here we report the demonstration of a mid-infrared DFG laser
sum of the squares of the errors A  KC, and the calculated C will
system for the simultaneous detection of different hydrocarbon
then be the best least-squares solution for equation (5).
species. Distinct from previous reports, the broad scanning range
The Euclidean norm of the errors can be written as:
will largely increase the number of measurable species. Fifteen dif-
ferent hydrocarbons covering major hydrocarbon groups were kA  KCk22 ¼ C T K T KC  2AT KC þ AT A ð6Þ
scanned. The measurements of the concentration dependence of
absorbance for each species were performed to validate the accu- By setting the expression to zero and performing matrix multi-
racy of the system. Measurements from binary to trinary mixtures plications, the concentrations of components in the mixture were
were also conducted to confirm the capability of mixture detection. calculated by the following equation:
The mixture measurement spectra were analyzed using the inde-  1
pendent components analysis and the classical least-squares C ¼ KT K KT A ð7Þ
method.
More detailed information on this method and the derivation of
the equation was discussed in [37].
2. Theory and experimental method
2.2. Experimental method
2.1. Theory
Fig. 1 displays the outline of the laser system. Two seed lasers
TDLAS follows Beer-Lambert’s law. According to this law, when sources, an external cavity diode laser (HSL-1–10-40-ZZ-Z-T-P,
light permeates an absorption medium, the molecular concentra- Stantec co.), which served as the signal source, and a DFB laser
tion is proportional to the strength of the transmitted light. The (DFB laser was placed inside the DFG module), which served as
relationship between the number density of measured species
and the amount of light absorbed can be explained by the follow-
ing formula:

Ik X X
¼ exp fAk g ¼ expf ðni L Si;j ðTÞGv i;j Þg ð1Þ
I0 i j

where Ik and I0 represents incident and transmitted light intensity,

respectively. Ak is absorbance, ni is the number density of species i, L Fig. 1. DFG laser system.

2
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

the pump source, were combined and transferred to the DFG laser Table 1 summarized the chosen single molecules and their
conversion module (WD-3236/3250–001-A-A-E, NTT Electronics experimental conditions. Experiment conditions for mixture mea-
co.) to generate the idler laser beam. The beam of the DFB laser surements were listed in Table 2. Hydrocarbon molecules with
was amplified using an amplifier (FA1550QLS, NTT Electronics over five carbon atoms were measured in their vapor phase under
co.) before the combination with the signal source. The power of higher temperature conditions (423 K) using the vapor phase
the idler wave is at the level of lw and the sweep span of the laser experiment setup. Others were measured using the gas phase
system is over 130 nm with a sampling rate of 6 kHz. setup. All experiments were performed under ambient pressure.
Fig. 2 and is a schematic drawing of the experimental setup for When performing the experiments, the pure nitrogen spectra were
the measurement of hydrocarbons in their gas phase. Hydrocarbon first recorded as the baseline, then the hydrocarbon/nitrogen mix-
and nitrogen gases were metered through independent mass flow ture spectra were recorded to calculate absorption spectra using
controllers (FCST1005MZF-4 J2-F1L-N2-R4, Fujikin Co., Ltd.) before Beer-Lambert’s law. The measured signals were averaged over
mixing. The concentrations of hydrocarbon molecules were the 200 ms period for noise reduction. Fig. 4 depicts the averaged
adjusted by controlling the flow rate of hydrocarbon and nitrogen pure nitrogen and hydrocarbon/nitrogen spectra for one measure-
gases. Gas pipes were wrapped with electrical heating tapes that ment. The horizontal axis displays the idler wavelength and was
were connected to temperature controllers for temperature adjust- calibrated with the HITRAN database. The signal intensity in wave-
ment. Heat isolation materials were covered outside gas pipes for length longer than 3390 nm was relatively weak and the signal was
heat reservation. Orange blocks represent the parts of the setup less stable in the wavelength shorter than 3260 nm, so only the
that can be heated. The mixed gas flowed through an absorbing cell middle part of the spectra was used for analysis. The absorption
with a laser path length of 100 mm. The absorbing cell was placed spectra of gas and vapor phase molecules were measured at seven
inside an electrical heating furnace which was also connected to a and five discrete concentrations, respectively. Measurements were
temperature controller. Signals from the DFG laser were detected repeated three times for each concentration/temperature
by an infrared detector (PVI-3TE-5-1x1-T08-wBaF2-35, VIGO Sys- condition.
tem S.A.) with a built-in preamplifier model (MIP-DC-20 MFM4,
VIGO System S.A.). The signals were then recorded by an oscillo- 2.3. Sensor performance
scope with a 125 MHz bandwidth and a total recording time of
200 ms (PicoScope 5444B, Pico Technology). The resolution of the Common spectroscopic databases for hydrocarbon molecules
system is around 0.01 cm1. include HITRAN [39], HITEMP [40], PNNL [41] and others (cf.
Some heavy hydrocarbons were in the liquid phase at room Table 4 in [6]). These databases provide line parameters or mea-
temperature and need to be heated to the gas phase before mea- sured cross sections for common hydrocarbon molecules. Detailed
suring. Fig. 3 illustrates the outline of the experimental setup for line parameters for methane and ethylene were included in the
these hydrocarbons. Most parts of the system were the same as HITRAN database. For other molecules mentioned in this work,
the setup in Fig. 2. A single syringe infusion pump (KDS-100, ISIS only cross-section information was available. Previous reports for
Co., Ltd.) was used instead of gas cylinders. A syringe was placed all the fifteen hydrocarbon measurements on the 3 lm region were
above the pump, and the concentration of liquid hydrocarbons summarized in Table 3 [41–47]. As can be seen from the table,
was adjusted by changing the injection speed of the syringe. Liq- most of the molecules lack accurate models in the 3 lm. The mea-
uids were injected into a heating chamber to make sure it was fully sured cross-sections also have different measurement conditions.
vaporized before entering the absorbing cell and another cell was So, it is difficult to make a direct comparison between measured
used as a container at the exit to collect condensed hydrocarbon. spectra and modeled spectra for each species. For sensitivity esti-
Fifteen hydrocarbons were chosen to cover various hydrocar- mation, the ‘per point’ minimum detectable absorption (MDApp)
bon classes and the different number of carbon atoms ranging from was calculated.
C1 to C8. The straight alkanes studied were methane (CH4), ethane The minimum detectable absorption (MDA) is a value cited by
(C2H6), propane (C3H8), n-butane (C4H10), n-pentane (C5H12), n- many researchers to evaluate the sensitivity performance of the
hexane (C6H14), and n-heptane (C7H16). Two straight alkenes cho- system. The MDApp was chosen since this definition is independent
sen were ethene (C2H4) and propene (C3H6). Benzene (C6H6), of scan characteristics and can be used to compare with other
toluene (C7H8), and m-xylene (C8H10) were picked for aromatics. instruments. The MDApp can be expressed as:
Three species belonging to other classes were methylcyclohexane  
DP 1
(C7H14), diisobutylene (C8H16), and isooctane (C8H18), which are MDApp ¼ pffiffiffiffiffiffiffiffi ð8Þ
P n BW
common components of gasoline surrogate fuels [38].

Fig. 2. Experimental Setup for gas phase molecules.

3
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Fig. 3. Experimental Setup for vapor phase molecules.

Table 1
Experimental conditions for single species.

Species Molecular Class Concentration Temperature Measurement Phase

Methane (CH4) Straight alkane 0.12%-0.58% 293–723 K Gas
Ethene (C2H4) Straight alkene 0.09%-0.99% 293 K Gas
Ethane (C2H6) Straight alkane 0.05%-0.20% 293 K Gas
Propene (C3H6) Straight alkene 0.09%-0.98% 293 K Gas
Propane (C3H8) Straight alkane 0.08%-0.39% 293 K Gas
n-Butane (C4H10) Straight alkane 0.06%-0.49% 293 K Gas
n-Pentane (C5H12) Straight alkane 0.42%-1.26% 423 K Vapor
n-Hexane (C6H14) Straight alkane 0.37%-1.10% 423 K Vapor
Methylcyclohexane Alkylcycloalkanes 1.14%-1.88% 423 K Vapor
(C7H14)
n-Heptane (C7H16) Straight alkane 0.33%-0.99% 423 K Vapor
Diisobutylene (C8H16) Branched alkene 0.31%-0.93% 423 K Vapor
Isooctane (C8H18) Branched alkane 0.30%-0.88% 423 K Vapor
Benzene (C6H6) Aromatic 1.09%-2.15% 423 K Vapor
Toluene (C7H8) Aromatic 1.36%-2.24% 423 K Vapor
m-Xylene (C8H10) Aromatic 0.79%-1.56% 423 K Vapor

Table 2 cheap components. A relatively good performance on sensitivity,

Experimental conditions for mixture measurements. scan rate, and resolution can be achieved when compared to the
Mixture Concentration Temperature above-listed devices on hydrocarbon measurement. At the same
Components time, the system can be applied to the real-time measurement per-
CH4 and C3H8 CH4: 0.12%-0.59%; C3H8: 0.20% 293 K formed on-site. Another merit is the calibration-free feature (cf.
CH4: 0.24%; C3H8: 0.1%-0.39% 293 K Fig. 6.7 in [3]). This feature ensures the proposed system to be
C2H4 and C3H6 C2H4: 0.12%-0.59%; C3H6: 0.40% 293 K unaffected by dust or small vibrations.
C2H4: 0.25%; C3H6:0.15%-0.74% 293 K
CH4, C2H4, and CH4:0.12%-0.59%; C2H4: 0.40%; 293 K
C3H8 C3H8:0.20%
3. Results and discussion
CH4: 0.20%; C2H4: 0.18%-0.77%; 293 K
C3H8:0.20%
CH4:0. 20%; C2H4: 0.4%; C3H8:0.10%- 293 K Due to the limited knowledge in accurate models for targeted
0.40% hydrocarbon, the measurement of the hydrocarbon spectra is nec-
essary to create a database that can be used for further analysis.
The results were discussed considering the different hydrocarbon
where DPP represents the limiting noise level of a measurement nor- classes and the number of carbon atoms.
malized by the total power, and BW is the sampling bandwidth. The CLS method relies on the linearity between concentration
To determine the noise level, a comparison between one mea- and absorbance, so concentration dependence experiments were
sured nitrogen spectrum and a polynomial fit to the data was carried out for each species to confirm their linear relationship.
made. The result was shown in Fig. 5. The mean square error of Based on pure hydrocarbon spectra, the performance of the system
MSE = 0.125% was obtained by calculating the square deviation of on mixture detection using the ICA and CLS method was examined.
each point. This value was used as DPP and a sampling rate of
6 kHz yields a value of MDApp = 2.29  10-5Hz1/2. Due to the high 3.1. Single species measurement
sampling rate, the minimum detectable absorption of the proposed
system is competitive compare to other broadband spectroscopic Because of the different composition of atoms and molecular
techniques (cf. Table 5 in [6]). structure, absorption of hydrocarbons is caused by different vibra-
Different spectrometers have trade-offs between sensitivity, tional modes and thus makes each molecule has a unique absorp-
path length, spectral resolution, and calibration. The proposed sys- tion pattern, which can be used to identify its kind. Although the
tem is a portable device system with lightweight and relatively number of carbon atoms will cause a bigger change in absorption
4
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Fig. 4. averaged spectra of CH4 and N2.

Table 3
List of hydrocarbon spectra in the 3 lm region reported by other researchers.

Species Data Type Resolution (cm1) Laser approach Reference

-7 -6
Methane (CH4) Line parameters 10 -10 DFG [42]
Ethene (C2H4) Line parameters 0.02 FTIR [43]
Ethane (C2H6) Cross sections 0.004 FTIR [44]
Propene (C3H6) Cross sections 0.112 FTIR [41]
Propane (C3H8) Cross sections 0.015 FTIR [45]
n-Butane (C4H10) Cross sections 0.112 FTIR [41]
n-Pentane (C5H12) Cross sections 0.1 FTIR [46]
n-Hexane (C6H14) Cross sections 0.112 FTIR [41]
Methylcyclohexane Cross sections / Not reported
(C7H14)
n-Heptane (C7H16) Cross sections 0.1 FTIR [46]
Diisobutylene (C8H16) Cross sections / Not reported
Isooctane (C8H18) Cross sections 0.1 FTIR [46]
Benzene (C6H6) Cross sections 0.112 FTIR [47]
Toluene (C7H8) Cross sections 0.1 FTIR [46]
m-Xylene (C8H10) Cross sections 0.1 FTIR [46]

spectra comparing to the difference in molecular structures, hydro- sity could introduce more noise and the absorbance is easier to
carbons that belong to the same group also tend to have similar fluctuate. Beer-Lambert’s law also may not be obeyed at higher
absorption shapes, especially when the number of carbon atoms absorbance. On the contrary, the R2 value in 3356.8 nm was the
was higher. highest where the signal intensity is stronger, and the line strength
Fig. 6(a) displayed the spectra of CH4. It has only one carbon is weaker. The absorption spectra of CH4 were also measured at
atom, which means the structure is relatively simple, and the five discrete temperatures between 293 K and 723 K and the
absorption lines are sharper compared to heavier molecules. results were shown in Fig. 6(c). The spectra are dependent on tem-
Fig. 6(b) showed the concentration dependence result at one wave- perature nonlinearly. The temperature variation measurement also
length. The X-axis is the concentration of CH4 gas, and the Y-axis is confirmed the capability of the system on high-temperature mea-
the absorbance at the peak near 3280.7 nm. A linear fitting line was surements under 723 K.
plotted to assess the linearity between concentration and absor- The spectra of other gas-phase molecules, namely C2H4, C2H6,
bance. For each species, five different wavelengths were picked C3H6, C3H8, and C4H10 were depicted in Fig. 7. C2H6 also has sharp
to calculate the concentration dependence on absorbance. The absorption peaks while C2H4 and C3H6 these two straight alkenes
selected wavelengths were to keep a relatively constant gap, and have more evenly distributed absorption in the scanning wave-
the absorption peaks were picked over an integer wavelength if length. The R2 values of the C3H6 result were smaller which was
peaks exist. The coefficient of determination, which is denoted as due to the weak absorption of the molecule. The lowest R2 value
R2, is a good indicator for the linear regression results. The selected was in 3310.0 nm where the highest absorbance was around 0.1
wavelengths, minimum and maximum R2 values for the concentra- under the current concentration range. C3H8 and C4H10 share sim-
tion dependence results of each species were summarized in ilar absorption spectra, which is the result of the CH3 asymmetric
Table 4. The high R2 values of the CH4 concentration dependence vibrational mode (for C4H10, the CH2 asymmetric mode also has
result show good linearity between concentration and absorbance. an impact, probably due to the equal number of CH2 and CH3
Several factors could affect the results. As for the CH4 results, the groups [48]).
lowest R2 value appeared in 3313.7 nm where the signal intensity Fig. 8 depicted the spectra of all the vapor phase molecules.
is weaker, and the line strength is stronger. A weaker signal inten- C5H12, C6H14, C7H14, and C7H16 are all aliphatic hydrocarbons with

5
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Table 4
Selected wavelength and R2 values for single species.

Species Selected Wavelength (nm) Minimum R2 Maximum R2

CH4 3280.7, 3313.7, 3334.4, 3356.8, 3379.9 0.9991 0.9999
C2H4 3267.1, 3284.6, 3346.2, 3375.5, 3314.7 0.9994 0.9997
C2H6 3328.9, 3336.8, 3348.1, 3359.3, 3373.4 0.9994 0.9999
C3H6 3310.0, 3330.7, 3343.1, 3361.3, 3384.7 0.9979 0.9991
C3H8 3330.0, 3345.0, 3369.7, 3376.6, 3385.0 0.9991 0.9999
C4H10 3330.0, 3345.0, 3362.4, 3371.2, 3385.0 0.9998 0.9999
C5H12 3340.0, 3350.0, 3360.0, 3372.4, 3385.0 0.9982 0.9986
C6H14 3340.0, 3350.0, 3360.0, 3373.3, 3385.0 0.9990 0.9992
C7H14 3350.0, 3360.0, 3370.0, 3381.6, 3385.0 0.9969 0.9986
C7H16 3340.0, 3350.0, 3360.0, 3370.0, 3380.0 0.9994 0.9996
C8H16 3340.0, 3350.0, 3360.0, 3370.0, 3385.0 0.9987 0.9996
C8H18 3340.0, 3350.0, 3360.0, 3370.0, 3385.0 0.9978 0.9992
C6H6 3265.0, 3275.0, 3285.0, 3295.0, 3305.0 0.9963 0.9978
C7H8 3265.0, 3275.0, 3285.0, 3295.0, 3305.0 0.9985 0.9990
C8H10 3275.0, 3290.0, 3305.0, 3320.0, 3335.0 0.9963 0.9976

Fig. 5. Measured spectrum and polynomial fit to the data. (a) Absorption spectra of CH4 (b) Concentration dependence of absorbance (c) Temperature dependence of
absorbance.

more than five carbon atoms. From the spectra, isolated narrow uncertainty of most vapor phase molecule results was larger than
absorption peaks cannot be found due to the complex structure light molecules. Despite the reasons that may affect the results
of polyatomic molecules. C8H16 and C8H18 only have blunted mentioned before, the difference in experimental setup could also
curves and indicates the limitation of the measurement system. have an impact, since the control of gas flow was different from the
C6H6, C7H8, and C8H10 belong to the aromatic group. Although their liquid.
absorption spectra are also the result of C-H stretching, unlike the From the results for each measured species, clear linearity
alkyl C-H stretching, the aromatic stretching is in a shorter wave- between concentration and absorbance can be seen from the cali-
length region (below 3320 nm), while the higher absorption area bration line. The overall R2 values were bigger than 0.996, indicat-
of heavier aliphatic compounds is over 3340 nm. Like other mole- ing good linear fitting results with small variance. The single
cules with a greater number of carbon atoms, these aromatics also species measurement results proved the system can measure
do not have very sharp peaks. The wavelengths selected for vapor hydrocarbons covering different molecular classes. The measured
phase molecules were the areas with the higher absorbance. The spectra of different hydrocarbons showed that the system has
the potential to separately detect hydrocarbons with up to eight
carbon atoms.

Table 5
Species identification results for mixtures.
3.2. Mixture measurement

Mixture Set Species Absorption Peak (nm) Correlation

Since hydrocarbon mixtures rather than pure species exist in
coefficients
actual conditions, the performance of the system on mixture detec-
CH4 and C3H8 CH4 3313.7 0.9502 tion should also be evaluated. Although the proposed system has a
C3H8 3369.7 0.9653
C2H4 and C3H6 C2H4 3346.2 0.9106
wider scanning range compared to other tunable diode lasers, not
C3H6 3384.7 0.8689 all hydrocarbons have isolated absorption peaks, as discussed in
CH4, C2H4, and C3H8 CH4 3313.7 0.9881 the previous section. When multiple components co-exist in the
C2H4 3346.2 0.8967 mixture, it is usually difficult to find an absorption peak that is free
C3H8 3369.7 0.9885
from the interference of other species.
6
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Fig. 6. Measurement results of CH4. (a) Absorption spectra of C2H4 (b) Absorption spectra of C2H6 (c) Absorption spectra of C3H6 (d) Absorption spectra of C3H8 (e) Absorption
spectra of C4H10.

The ICA method was applied for the identification of compo- The absorption spectra measured in each set were used for ICA
nents from the mixture and the CLS method was used for concen- decomposition and the components were identified through peak
tration calculations. By using the two methods, it is possible to searching and similarity comparison among separated spectra
identify different components within the mixture and determine and the spectral database. The separated spectra were displayed
their concentrations. First, the measured mixture spectra were in Fig. 9(b) and Fig. 10(b). All gases were successfully recognized
used as input, and then the components were separated using through absorption peaks, and further confirmation was achieved
the ICA method. Then the spectra of pure species measured at each by calculating the correlation coefficient r between the pure spec-
of their highest concentrations were used together with mixture trum and the separated spectrum. The correlation coefficient r
spectra to calculate the concentrations using the CLS method. equation is as follows:
The ‘‘actual” and ‘‘calculated” concentrations were compared and P 
  

assessed using both the R2 and the root mean square error of pre- ½ xi  x  y i  y 
r ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi; i ¼ 1; 2; 3    ð8Þ
P P
diction (RMSEP). The term ‘‘actual” and ‘‘calculated” refer to the
 2  2
known concentration of selected components in the mixture and ½ xi  x  ½ yi  y
the value calculated using the CLS method, respectively. The ‘‘ac-
tual” concentrations were calculated using the readings from the where  and y represent the data points in the two spectra being
flow meters. compared and r greater than 0.9 indicates the two spectra were well
The spectra of several mixtures with certain concentrations matched according to the previous study [49]. In the two experi-
were recorded and analyzed. The chosen molecules could cover ments sets, all values of r were greater than 0.86, showing good sep-
different absorption shapes. CH4 and C2H4 have narrow sharp aration results using the ICA decomposition method. The separation
peaks, C3H6 has an absorption in the whole scanning range, C3H8 results for the CH4 and C3H8 set were better than the C2H4 and C3H6
has blunted curve in the longer wavelength region. The chosen set, which can be explained by the complex absorption structure
mixture sets were divided as partially and fully overlapping with and overlapped absorption bands. The identification results were
each other. summarized in Table 5.
Two sets of binary mixtures were chosen for validation, one set After species identification, quantitative analysis was per-
was CH4 and C3H8 and another was C2H4 and C3H6. The first set was formed using the CLS method to determine the variation in concen-
partially overlapped in the detection wavelength range, and the tration. The calculation results using the CLS method were
second set was completely overlapped. Two experiments were car- displayed in Fig. 9(c) and Fig. 9(d). The R2 values obtained for
ried out for each set. In both experiments, the concentration of one CH4 and C3H8 respectively were 0.9992 and 0.9999, RMSEP values
species remains constant while the concentration of the other were 0.012% and 0.008%. Both values indicate good linearity and a
varies. small variance between actual and calculated concentrations.
7
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Fig. 7. Absorption spectra of gas phase hydrocarbons. (a) Absorption spectra of C5H12 (b) Absorption spectra of C6H14 (c) Absorption spectra of C7H14 (d) Absorption spectra of
C7H16 (e) Absorption spectra of C8H16 (f) Absorption spectra of C8H18 (g) Absorption spectra of C6H6 (h) Absorption spectra of C7H8 (i) Absorption spectra of C8H10.

The spectra and results for C2H4 and C3H6 were shown in Fig. 10. C3H8 were 0.988, 0.896, and 0.989, respectively as shown in
The R2 values for C2H4 and C3H6 were 0.9981 and 0.9985, RMSEP Fig. 11(b). The C2H4 correlation was the lowest, which was consis-
values were 0.033% and 0.029%. Both values also indicate good fit- tent with the binary measurement results.
ting results although the results for CH4 and C3H8 were better. The The calculated concentration results were illustrated in Fig. 11
overlapping between these two molecules might influence the (c), Fig. 11(d), and Fig. 11(e). For CH4, the R2 and RMSEP values
mixture spectra that worsen the fitting result. The concentration were 0.9994 and 0.020%, which was a little worse than the CH4
dependence results for C3H6 showed more uncertainty since the and C3H8 mixture measurement result. For C2H4 and C3H8, the
calculation used the pure spectra of single species, this could also results were almost identical to binary mixture detection, with
affect the mixture detection results. the R2 values 0.9997 and 0.9999, the RMSEP values 0.039% and
By expanding the experiment to trinary mixtures, the perfor- 0.004% for each species. With the presence of another interferent
mance of the system was further validated. Fig. 11(a) showed the in mixtures, only small changes were observed compared to binary
absorption spectra of the CH4, C2H4, and C3H8 mixture. The exper- mixture measurements. Overall, the high R2 and low RMSEP values
iments were like binary measurements while the concentration of proved the CLS method is reliable in predicting the concentration
one species change and the other two remain the same. Peak of mixture components.
searching was first performed for the separated spectra and all The experiments being carried out were able to prove the pro-
three species were identified. The r values for CH4, C2H4, and posed system is capable of mixture hydrocarbon measurements.

8
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Fig. 8. Absorption spectra of vapor phase hydrocarbons. (a) Absorption spectra of CH4 and C3H8 mixture (b) Separated results of CH4 and C3H8 mixture (c) Actual and
calculated concentration of CH4 (d) Actual and calculated concentration of C3H8.

9
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Fig. 10. Measurement results of C2H4 and C3H6. (a) Absorption spectra of CH4, C2H4
Fig. 9. Measurement result of CH4 and C3H8 mixture. a) Absorption spectra of C2H4
and C3H8 mixture (b) Separated results of CH4, C2H4 and C3H8 mixture (c) Actual
and C3H6 mixture (b) Separated results of C2H4 and C3H6 mixture (c) Actual and
and calculated concentration of CH4 (d) Actual and calculated concentration of C2H4
calculated concentration of C2H4 (d) Actual and calculated concentration of C3H6.
(e) Actual and calculated concentration of C3H8.

10
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

concentration of different components without the need to find

isolated absorption peaks.

4. 4.Conclusion

This study presented a DFG laser system operating near the

3.3 lm region with a wide scanning range of over 130 nm for
hydrocarbon detection. The system was validated regarding its
capability of both pure and mixture measurement. Based on the
results of this study, conclusions were summarized as follows.

1) The absorption spectra of fifteen different hydrocarbon

molecules ranging from C1 to C8 were probed. Results of
the concentration dependence of absorbance were obtained
for each molecule. Linear dependence results with the R2
values larger than 0.997 demonstrated the system was able
to measure hydrocarbons covering different molecular
classes accurately. The measurement of methane at different
temperatures confirmed the nonlinear relationship between
temperature and absorbance.
2) Mixtures of hydrocarbon gases measurements from binary
to trinary mixtures were performed to validate the perfor-
mance of the system on mixture detection, which is impor-
tant in actual applications. The values of r larger than 0.86
showed that the ICA algorithm can achieve mixture separa-
tion under the given conditions. The R2 values higher than
0.998 and RMSEP values smaller than 0.04% proved that
the CLS method can be applied to predict the concentrations
of target species in the mixture by using the spectra of pure
species. These methods can provide good separation and
determination of the concentration on binary and trinary
mixtures.

The proposed system is a good candidate for industrial hydro-

carbon detection by overcoming the limitation of other tunable
diode lasers designed for hydrocarbon measurement while main-
taining its advantages. Further studies are still necessary for more
complicated mixtures in actual situations, more species, and inter-
ferents than designed conditions in this study. Other works include
the acquisition of hydrocarbons under different temperature con-
ditions and the validation of actual processes.

CRediT authorship contribution statement

Qiming Wang: Writing – review & editing. Zhenzhen Wang:

Writing – review & editing. Takahiro Kamimoto: Investigation,
Resources. Yoshihiro Deguchi: Methodology, Supervision. Shengli
Cao: . Du Wen: Data curation, Writing – review & editing. Daichi
Takahara: Investigation, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared
to influence the work reported in this paper.

References

[1] R.K. Hanson, P.K. Falcone, Temperature measurement technique for high-
temperature gases using a tunable diode laser, Appl. Opt. 17 (16) (1978) 2477,
Fig. 11. Measurement results of CH4, C2H4 and C3H8 mixture. https://doi.org/10.1364/ao.17.002477.
[2] M. Lackner, Tunable diode laser absorption spectroscopy (TDLAS) in the
process industries - A review, Rev. Chem. Eng. 23 (2) (2007) 65–147, https://
doi.org/10.1515/REVCE.2007.23.2.65.
The results suggest the ICA and the CLS method can be used to [3] Y. Deguchi, Industrial Applications of Laser Diagnostics, CRC Press, 2016.
identify species in the mixture and quantitatively determine the [4] M.A. Bolshov, Y.A. Kuritsyn, Y.V. Romanovskii, ‘‘Tunable diode laser
spectroscopy as a technique for combustion diagnostics”, Spectrochim, Acta -

11
Q. Wang, Z. Wang, T. Kamimoto et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 265 (2022) 120333

Part B At. Spectrosc. 106 (2015) 45–66, https://doi.org/10.1016/j. [26] D. Richter, P. Weibring, J.G. Walega, A. Fried, S.M. Spuler, M.S. Taubman,
sab.2015.01.010. Compact highly sensitive multi-species airborne mid-IR spectrometer, Appl.
[5] G. A. Eiceman, H. H. Hill, and J. Gardea-Torresdey, ‘‘Gas Chromatography,” Anal. Phys. B Lasers Opt. 119 (1) (2015) 119–131, https://doi.org/10.1007/s00340-
Chem., vol. 70, no. 12, 1998, 10.1021/a1980016l. 015-6038-8.
[6] K.C. Cossel, E.M. Waxman, I.A. Finneran, G.A. Blake, J. Ye, N.R. Newbury, Gas- [27] J.H. Van Helden, G. Hancock, R. Peverall, G.A.D. Ritchie, A 3 lm difference
phase broadband spectroscopy using active sources: progress, status, and frequency laser source for probing hydrocarbon plasmas, J. Phys. D. Appl. Phys.
applications [Invited], J. Opt. Soc. Am. B 34 (1) (2017) 104, https://doi.org/ 44 (12) (2011), https://doi.org/10.1088/0022-3727/44/12/125202 125202.
10.1364/josab.34.000104. [28] E.-T. Es-sebbar, M. Alrefae, A. Farooq, Infrared cross-sections and integrated
[7] S. H. Pyun, J. Cho, D. F. Davidson, and R. K. Hanson, ‘‘Interference-free mid-IR band intensities of propylene: Temperature-dependent studies, J. Quant.
laser absorption detection of methane,” Meas. Sci. Technol., vol. 22, no. 2, 2011, Spectrosc. Radiat. Transf. 133 (2014) 559–569, https://doi.org/10.1016/j.
10.1088/0957-0233/22/2/025303. jqsrt.2013.09.019.
[8] K. Tanaka, K. Akishima, M. Sekita, K. Tonokura, M. Konno, Measurement of [29] J. Cousin, W. Chen, D. Bigourd, M. Fourmentin, S. Kassi, Telecom-grade fiber
ethylene in combustion exhaust using a 3.3-lm distributed feedback laser-based difference-frequency generation and ppb-level detection of
interband cascade laser with wavelength modulation spectroscopy, Appl. benzene vapor in air around 3 lm, Appl. Phys. B Lasers Opt. 97 (4) (2009)
Phys. B Lasers Opt. 123 (8) (2017) 1–8, https://doi.org/10.1007/s00340-017- 919–929, https://doi.org/10.1007/s00340-009-3669-7.
6798-4. [30] M.K. Shakfa et al., A mid-infrared diagnostic for benzene using a tunable
[9] R.S.M. Chrystie, E.F. Nasir, A. Farooq, Propene concentration sensing for difference-frequency-generation laser, Proc. Combust. Inst. 38 (1) (2021)
combustion gases using quantum-cascade laser absorption near 11 lm, Appl. 1787–1796, https://doi.org/10.1016/j.proci.2020.06.382.
Phys. B Lasers Opt. 120 (2) (2015) 317–327, https://doi.org/10.1007/s00340- [31] A.E. Klingbeil, J.M. Porter, J.B. Jeffries, R.K. Hanson, Two-wavelength mid-IR
015-6139-4. absorption diagnostic for simultaneous measurement of temperature and
[10] R. Sur, S. Wang, K. Sun, D.F. Davidson, J.B. Jeffries, R.K. Hanson, High-sensitivity hydrocarbon fuel concentration, Proc. Combust. Inst. 32 (1) (2009) 821–829,
interference-free diagnostic for measurement of methane in shock tubes, J. https://doi.org/10.1016/j.proci.2008.06.058.
Quant. Spectrosc. Radiat. Transf. 156 (2015) 80–87, https://doi.org/10.1016/j. [32] A.E. Klingbeil, J.B. Jeffries, R.K. Hanson, Tunable mid-IR laser absorption sensor
jqsrt.2015.01.023. for time-resolved hydrocarbon fuel measurements, Proc. Combust. Inst. 31 (1)
[11] K. Krzempek, R. Lewicki, L. Nähle, M. Fischer, J. Koeth, S. Belahsene, Y. (2007) 807–815, https://doi.org/10.1016/j.proci.2006.07.228.
Rouillard, L. Worschech, F.K. Tittel, Continuous wave, distributed feedback [33] Y.B. Monakhova, S.A. Astakhov, A. Kraskov, S.P. Mushtakova, Independent
diode laser based sensor for trace-gas detection of ethane, Appl. Phys. B Lasers components in spectroscopic analysis of complex mixtures, Chemom. Intell.
Opt. 106 (2) (2012) 251–255, https://doi.org/10.1007/s00340-011-4857-9. Lab. Syst. 103 (2) (2010) 108–115, https://doi.org/10.1016/j.
[12] P. Kluczynski, S. Lundqvist, S. Belahsene, Y. Rouillard, L. Nähle, M. Fischer, J. chemolab.2010.05.023.
Koeth, Detection of propane using tunable diode laser spectroscopy at 3.37 [34] G. Wang, Q. Ding, Z. Hou, Independent component analysis and its applications
lm, Appl. Phys. B Lasers Opt. 108 (1) (2012) 183–188, https://doi.org/10.1007/ in signal processing for analytical chemistry, TrAC - Trends Anal. Chem. 27 (4)
s00340-012-5054-1. (2008) 368–376, https://doi.org/10.1016/j.trac.2008.01.009.
[13] N. Lang, U. Macherius, M. Wiese, H. Zimmermann, J. Röpcke, J.H. van Helden, [35] X. Yu, Y. Zhang, G. Yin, N. Zhao, X. Xiao, C. Lu, Y. Gao, W. Zhang, Discrimination
Sensitive CH_4 detection applying quantum cascade laser based optical of three dimensional fluorescence spectra based on wavelet analysis and
feedback cavity-enhanced absorption spectroscopy, Opt. Express 24 (6) independent component analysis, Spectrochim. Acta - Part A Mol. Biomol.
(2016) A536, https://doi.org/10.1364/oe.24.00a536. Spectrosc. 124 (2014) 52–58, https://doi.org/10.1016/j.saa.2013.12.033.
[14] Y. Cao et al., Simultaneous atmospheric nitrous oxide, methane and water [36] A. Hyvärinen, E. Oja, A Fast Fixed-Point Algorithm for Independent Component
vapor detection with a single continuous wave quantum cascade laser, Opt. Analysis, Neural Comput. 9 (7) (1997) 1483–1492, https://doi.org/10.1162/
Express 23 (3) (2015) 2121, https://doi.org/10.1364/oe.23.002121. neco.1997.9.7.1483.
[15] L. Dong, F.K. Tittel, C. Li, N.P. Sanchez, H. Wu, C. Zheng, Y. Yu, A. Sampaolo, R.J. [37] R. Kramer, Chemometric Techniques for Quantitative Analysis, CRC Press,
Griffin, Compact TDLAS based sensor design using interband cascade lasers for 1998.
mid-IR trace gas sensing, Opt. Express 24 (6) (2016) A528, https://doi.org/ [38] S.M. Sarathy, A. Farooq, G.T. Kalghatgi, Recent progress in gasoline surrogate
10.1364/OE.24.00A528. fuels, Prog. Energy Combust. Sci. 65 (2018) 67–108, https://doi.org/10.1016/j.
[16] S. Wang, T. Parise, S.E. Johnson, D.F. Davidson, R.K. Hanson, A new diagnostic pecs.2017.09.004.
for hydrocarbon fuels using 3.41-mm diode laser absorption, Combust. Flame [39] I.E. Gordon et al., The HITRAN2016 molecular spectroscopic database, J. Quant.
186 (2017) 129–139, https://doi.org/10.1016/j.combustflame.2017.07.026. Spectrosc. Radiat. Transf. 203 (2017) 3–69, https://doi.org/10.1016/j.
[17] A. Sampaolo, S. Csutak, P. Patimisco, M. Giglio, G. Menduni, V. Passaro, F.K. jqsrt.2017.06.038.
Tittel, M. Deffenbaugh, V. Spagnolo, Methane, ethane and propane detection [40] L.S. Rothman et al., HITEMP, the high-temperature molecular spectroscopic
using a compact quartz enhanced photoacoustic sensor and a single interband database, J. Quant. Spectrosc. Radiat. Transf. 111 (15) (2010) 2139–2150,
cascade laser, Sensors Actuators, B Chem. 282 (2019) 952–960, https://doi.org/ https://doi.org/10.1016/j.jqsrt.2010.05.001.
10.1016/j.snb.2018.11.132. [41] T.J. Johnson, R.L. Sams, S.W. Sharpe, in: in Chemical and Biological Point Sensors
[18] D. Richter, P. Weibring, Ultra-high precision mid-IR spectrometer I: Design and for Homeland Defense, 2004, p. 159, https://doi.org/10.1117/12.515604.
analysis of an optical fiber pumped difference-frequency generation source, [42] M. Abe, K. Iwakuni, S. Okubo, H. Sasada, Accurate transition frequency list of
Appl. Phys. B Lasers Opt. 82 (3) (2006) 479–486, https://doi.org/10.1007/ the m_3 band of methane from sub-Doppler resolution comb-referenced
s00340-005-2072-2. spectroscopy, J. Opt. Soc. Am. B 30 (4) (2013) 1027, https://doi.org/10.1364/
[19] P. Weibring, D. Richter, A. Fried, J.G. Walega, C. Dyroff, Ultra-high-precision josab.30.001027.
mid-IR spectrometer II: System description and spectroscopic performance, [43] M. Bach, R. Georges, M. Hepp, M. Herman, Slit-jet Fourier transform infrared
Appl. Phys. B Lasers Opt. 85 (2–3) (2006) 207–218, https://doi.org/10.1007/ spectroscopy in 12C2H4: Cold and hot bands near 3000 cm-1, Chem. Phys.
s00340-006-2300-4. Lett. 294 (6) (1998) 533–537, https://doi.org/10.1016/S0009-2614(98)00889-
[20] W. Chen et al., Continuous-wave mid-infrared laser sources based on 6.
difference frequency generation, Comptes Rendus Phys. 8 (10) (2007) 1129– [44] J.J. Harrison, N.D.C. Allen, P.F. Bernath, Infrared absorption cross sections for
1150, https://doi.org/10.1016/j.crhy.2007.09.011. ethane (C2H6) in the 3 lm region, J. Quant. Spectrosc. Radiat. Transf. 111 (3)
[21] K.P. Petrov, S. Waltman, E.J. Dlugokencky, M. Arbore, M.M. Fejer, F.K. Tittel, L. (2010) 357–363, https://doi.org/10.1016/j.jqsrt.2009.09.010.
W. Hollberg, Precise measurement of methane in air using diode-pumped 3.4- [45] J.J. Harrison, P.F. Bernath, Infrared absorption cross sections for propane
lm difference-frequency generation in PPLN, Appl. Phys. B Lasers Opt. 64 (5) (C3H8) in the 3lm region, J. Quant. Spectrosc. Radiat. Transf. 111 (9) (2010)
(1997) 567–572, https://doi.org/10.1007/s003400050216. 1282–1288, https://doi.org/10.1016/j.jqsrt.2009.11.027.
[22] D. Richter, D.G. Lancaster, R.F. Curl, W. Neu, F.K. Tittel, Compact mid-infrared [46] A.E. Klingbeil, J.B. Jeffries, R.K. Hanson, Temperature-dependent mid-IR
trace gas sensor based on difference-frequency generation of two diode lasers absorption spectra of gaseous hydrocarbons, J. Quant. Spectrosc. Radiat.
in periodically poled LiNbO3, Appl. Phys. B Lasers Opt. 67 (3) (1998) 347–350, Transf. 107 (3) (2007) 407–420, https://doi.org/10.1016/j.jqsrt.2007.03.004.
https://doi.org/10.1007/s003400050514. [47] C.P. Rinsland, V.M. Devi, T.A. Blake, R.L. Sams, S. Sharpe, L. Chiou, Quantitative
[23] U. Gustafsson, J. Sandsten, S. Svanberg, Simultaneous detection of methane, measurement of integrated band intensities of benzene vapor in the mid-
oxygen and water vapour utilising near-infrared diode lasers in conjunction infrared at 278, 298, and 323 K, J. Quant. Spectrosc. Radiat. Transf. 109 (15)
with difference-frequency generation, Appl. Phys. B Lasers Opt. 71 (6) (2000) (2008) 2511–2522, https://doi.org/10.1016/j.jqsrt.2008.04.007.
853–857, https://doi.org/10.1007/s003400000431. [48] L. Pele, J. Šebek, E.O. Potma, R. Benny Gerber, Raman and IR spectra of butane:
[24] K.E. Whittaker, L. Ciaffoni, G. Hancock, R. Peverall, G.A.D. Ritchie, A DFG-based Anharmonic calculations and interpretation of room temperature spectra,
cavity ring-down spectrometer for trace gas sensing in the mid-infrared, Appl. Chem. Phys. Lett. 515 (1–3) (2011) 7–12, https://doi.org/10.1016/j.
Phys. B Lasers Opt. 109 (2) (2012) 333–343, https://doi.org/10.1007/s00340- cplett.2011.09.015.
012-5150-2. [49] M. Mecozzi, M. Pietroletti, Y.B. Monakhova, FTIR spectroscopy supported by
[25] R. Grilli, L. Ciaffoni, G. Hancock, R. Peverall, G.A.D. Ritchie, A.J. Orr-Ewing, Mid- statistical techniques for the structural characterization of plastic debris in the
infrared ethene detection using difference frequency generation in a quasi- marine environment: Application to monitoring studies, Mar. Pollut. Bull. 106
phase-matched LiNbO3 waveguide, Appl. Opt. 48 (30) (2009) 5696–5703, (1–2) (2016) 155–161, https://doi.org/10.1016/j.marpolbul.2016.03.012.
https://doi.org/10.1364/AO.48.005696.

12